JP6572758B2 - Program and information processing apparatus - Google Patents

Description

  The present invention relates to a program and an information processing apparatus.

  Research on “complex networks”, which are huge and complex networks existing in the real world, is advancing and various analyzes of complex networks and services using the analysis results have been proposed. The complex network includes, for example, a WWW (World Wide Web) composed of a huge number of web page groups, the Internet, a social network, a document citation network, a user product network, a gene control network, a protein interaction network, and the like.

  In complex network science, a portion where links (also called edges) in a network are dense is called a “community” (see Non-Patent Documents 1 and 2 described later). A community is also considered to correspond to each functional module that constitutes a complex system expressed by a network. Various methods for extracting a community from a network such as a complex network have been proposed.

  Non-Patent Documents 1 and 2 are reviews on algorithms for detecting a community from a network. Of the community detection algorithms that have been proposed so far, notable ones (eg, modularity maximization, information compression, label propagation, clique percolation, block model) are all described in these documents.

  Non-Patent Documents 3 to 6 disclose a method for detecting a community including one seed node designated by a user or the like as a member. In the methods disclosed in these documents, a node is added to a community while tracing a node in the network starting from a designated seed node, and a community evaluation value is calculated by an evaluation function. The target community is obtained by optimizing the evaluation value.

  In Patent Document 1, in a network in which transition between nodes follows a Markov chain, a path having the highest possibility of following a path starting from a specified node is repeatedly calculated based on the transition probability of the Markov chain. The method to obtain is described.

  In Patent Document 2, in a network in which transitions between nodes follow a Markov chain, the computation processing of the personalized page rank algorithm, which has a high calculation load for obtaining a node group that matches a search condition from the user, is not applied to the entire network. A method for limiting to a part related to a search condition is described. In this method, the clustering of Markov chains on the network is learned using the learning data, and the cluster that matches the search condition is extracted from the node group corresponding to the search condition from the user and the learning result of the clustering, A personalized page rank algorithm is executed on a partial network composed of nodes belonging to the cluster.

  The information processing apparatus described in Patent Document 3 includes a log acquisition unit that acquires log information from a document DB (database), and a user who operates the document information as a node, and a user who operates the same document information. Network generating means for creating a network structure by establishing a link, and community extracting means for extracting a set of a plurality of users having a common relationship as a community from the network structure using a probability model based on the generation probability of the link A user index calculating means for calculating, as an index, an affiliation rate that is a probability distribution of users in the community or an influence degree that is an expected value of the network index of users in the community.

Newman, M.E.J.Communities, modules and large-scale structure in networks.Nature Physics 8, 25-31 (2011) Fortunato, S. Community detection in graphs.Physic Report 486, 75-174 (2010) Bagrow, J.P., Bollt, E.M .: Local method for detecting communities.Phys. Rev. E 72, 046108 (2005) Clauset, A .: Finding local community structure in networks. Phys. Rev. E 72, 026132 (2005) Lancichinetti, A., Fortunato, S., Kertesz, J .: Detecting the overlapping and hierarchical community structure in complex networks.New J. Phys. 11, 033015 (2009) Okamoto, H .: Local Detection of Communities vt Neural-Network Dynamics. LNCS 8131, 50--57 (2013)

JP 2013-41530 A JP2013-168127A Japanese Patent Laying-Open No. 2015-46102

  In a method of adding a node to a community while tracing a node in the network starting from a designated seed node, and obtaining a community having a high evaluation value, the user needs to designate the seed node. Such a method cannot be used when the user cannot identify the seed node or when the user cannot narrow down to one seed node.

  An object of the present invention is to enable extraction of a community in a network corresponding to a user query even when the user cannot identify a single node in the network as a seed node.

  According to the first aspect of the present invention, there is provided an evaluation value calculating means for calculating an evaluation value indicating a degree of conformity with respect to a query of each node in the network, and the first component of the evaluation value of each node is obtained from the node. According to a calculation rule that a transition is made to any one of linked nodes according to an outgoing link, and the second component of the evaluation value of each node transits to another node regardless of the link exiting from the node, The second component transition calculation means for calculating an evaluation value of each node after the transition, wherein, in the calculation rule, the second component is obtained from a node having an evaluation value within a first range close to 0. More transitions to nodes having an evaluation value higher than the first range, and the proportion of the first component in the evaluation value of the node increases monotonically as the evaluation value increases That the transition calculation means, based on the evaluation value after the transition of the calculated each node of the transition computing means, a program for causing the specifying means functions as, identifying one or more nodes that correspond to the query.

  According to a second aspect of the present invention, the specifying unit sets, as a community corresponding to the query, a node group in which an evaluation value after transition of each node calculated by the transition calculating unit is larger than a predetermined threshold value. 2. The program according to claim 1, wherein the program is specified.

  The invention according to claim 3 is characterized in that the specifying unit generates a list showing the specified node group in descending order of evaluation values after transition of the nodes calculated by the transition calculating unit. A program according to claim 1 or 2.

  The invention according to claim 4 is characterized in that, in the calculation rule, the ratio of the first component to the evaluation value of the node is 0 for a node having the evaluation value equal to or less than a threshold value. The program according to any one of 1 to 3.

  The invention according to claim 5 is an evaluation value calculation means for calculating an evaluation value indicating a degree of conformity to a query of each node in the network, and a first component of the evaluation value of each node is a link coming out of the node According to the calculation rule that the second component of the evaluation value of each node transits to another node regardless of the link exiting from the node. The second component transition calculation means for calculating the evaluation value of each of the nodes, wherein, in the calculation rule, the second component is more than the node having the evaluation value in the first range close to 0. Transition calculation with more transitions for nodes having an evaluation value higher than the first range, and the proportion of the first component in the evaluation value of the node monotonously increases as the evaluation value increases Stage and, based on the evaluation value after the transition of the calculated each node of the transition computing means, and identifying means for identifying one or more nodes that correspond to the query, an information processing apparatus having a.

  According to the first, second, or fifth aspect of the present invention, it is possible to extract a community in the network corresponding to a user query even when the user cannot identify a single node in the network as a seed node.

  According to the third aspect of the present invention, it is possible to provide a list of nodes having a high evaluation value for a query in consideration of not only the degree of matching of the node with respect to the query but also the link status between the nodes in the network.

  According to the invention of claim 4, compared to the case where the ratio of the first component corresponding to the evaluation value equal to or less than the threshold is set to a value larger than 0, the calculation in the transition calculating means includes the community corresponding to the query. The difference between a node and a node not included in the community can be made clearer.

It is a figure which shows an example of a function structure of the information processing apparatus of embodiment. It is a diagram illustrating an example of a graph showing the relationship between evaluation values p _n (t) and the link transition component f _n (t) therein. While there is a possibility that a false member that does not belong to the community may be extracted only by the degree of conformity to the query, the community that excludes the false member by repeatedly calculating the time evolution of the evaluation value in consideration of the link status of the network It is a figure for demonstrating that is extracted. A diagram showing another example of a graph showing the relationship between evaluation values p _n (t) and the link transition component f _n (t) therein is.

  With reference to FIG. 1, an apparatus configuration according to an embodiment of the present invention will be described.

The network information storage unit 10 is a storage device that stores information (hereinafter referred to as “network information”) representing a network (for example, WWW) to be processed. The network information includes information indicating the structure of the network, that is, connection relationship information indicating a link connection relationship between nodes in the network. This connection relation information is expressed in the form of an adjacency matrix, for example. As is well known, the adjacency matrix A _nm is a matrix in which A _nm = 1 if a link is established from the node m to the node n, and A _nm = 0 otherwise. The network connection relation information may be information expressed in the form of a transition matrix T _nm . The transition matrix T _nm is a matrix indicating the probability that an agent on the node m follows the link from the node m to the node n and transitions to the node n. If there is no link from node m to node n, the value of T _nm is zero. When attention is paid to a certain node m, the sum of T _nm over all nodes n in the network is “1”.

  The network information may include attribute information of each node. The attribute information of a node includes one or more attribute values possessed by the entity represented by the node. For example, when a node represents a web page, content data of the web page (description of HTML representing the page) is one of the attribute information of the node. If the node represents a paper, the attribute information of the node may include the content (text) of the paper, the author name, the journal name, and the like. If the target network is a social network, each node corresponds to a person. In this case, the attribute information of the node includes the name, age, gender, occupation, organization, etc. of the person corresponding to the node. Information may be included.

  When the information on the network to be processed exists outside the apparatus, the network information acquisition unit 12 acquires the network information. For example, when the target network is the WWW, the network information acquisition unit 12 collects web page groups by following links by a so-called crawler, and the connection (link) relationship between nodes (web pages) and each of the collected information. Extract node attribute information. When a social network managed by an SNS (social network service) is targeted, the network information acquisition unit 12 acquires information on the social network by using an interface published by the SNS. . The network information acquired by the network information acquisition unit 12 is stored in the network information storage unit 10.

  The query receiving unit 14 receives a query (inquiry) for the target network from the user. The query is a condition for narrowing down a node group in the target network that is suitable for the user's intention. The format of the query is not particularly limited.

  For example, the query may be a search condition related to one or more attributes of the node. When the node corresponds to a document such as a web page, a query may be a logical expression including a keyword (search term) or a plurality of keywords. Moreover, it is good also considering the combination of the conditions regarding a some attribute like an author and a publication journal name as a query.

  Further, the query may directly specify one or more specific nodes among the node group in the target network. For example, if you want to find a community of junior high school alumni from a large number of people (nodes) that make up a social network, you can specify the number of people you are familiar with as alumni and query the combination of those people This is the case.

  The initial evaluation value calculation unit 16 calculates an initial value (initial evaluation value) of an evaluation value indicating the degree of suitability of each node for the query received by the query reception unit 14. When the query is a search condition, the initial evaluation value may be calculated by a calculation method of a search algorithm that calculates a fitness (score) for the search condition. For example, based on the attribute information of each node stored in the network information storage unit 10, the degree to which each node meets the search condition may be calculated. Moreover, you may use what normalized the matching degree which the search algorithm calculated | required (for example, the process which makes the sum total of the evaluation value of all the nodes become 1) as an evaluation value. In the case of a query that directly specifies one or more nodes, the initial evaluation value of nodes other than the specified node is 0, and the initial evaluation value of each specified node is a positive value (for example, a node with 1 specified). Or a value divided by the number of In addition, an initial evaluation value of each node designated by the user may be input. The “evaluation value” of the node is a real number of 0 or more.

  The time evolution calculation unit 18 calculates the time evolution of the evaluation value of each node.

  A page rank algorithm is well known as an algorithm for calculating the time evolution of the evaluation value (page rank value) of each node in the network. The page rank algorithm is based on the Markov chain random walk analogy in which one agent walks at random while following links between nodes in the network, and a node with a higher probability that an agent exists has a higher page rank value. The page rank algorithm assumes that the agent on the node m selects one or more links coming out of the node m with equal probability and makes a transition to the link destination node, and calculates the time evolution of the evaluation value of the node group. A method is also known in which the strength of each link is obtained from information actually obtained, and the probability that the agent selects the link according to the strength of the link is known. The calculation of the time evolution of the node evaluation value by the conventional method such as the page rank algorithm is based on the premise that the transition between the nodes of the agent is performed through the link.

  On the other hand, in the calculation rule used in the time evolution calculation unit 18 of the present embodiment, the agent transition between nodes includes not only a transition component passing through a link on the network but also a link with respect to a node having a high evaluation value. It is assumed that components of transitions that jump directly without passing through may also be included. That is, the evaluation value of a node is divided into a first component and a second component, the first component transits to another node through a link exiting from the node, and the second component transits to another node regardless of the link. It shall be. The first component is called a “link transition component”, and the second component is called a “non-link transition component”. Of the evaluation value of a node, a link transition component can only transition to any one of the nodes linked to that node, but a non-link transition component can also transition to a node other than the link destination of that node.

  Further, according to the calculation rule of the time evolution calculation unit 18, the non-link transition component is likely to transition to a node having a high evaluation value (that is, a larger amount transitions). For example, the degree (probability) of transition of a non-link transition component of a certain node may monotonously increase as the evaluation value of the transition destination node increases. Looking at this from the opposite side, the portion derived from the transition of the non-link transition component of another node group (to the node) in the node evaluation value increases monotonously as the evaluation value of the node increases. It will be. By defining the non-link transition component in this way, the non-link transition component is likely to gather at a node having a high evaluation value. As an example, the transition destination of the non-link transition component may be limited to a node group whose evaluation value is equal to or greater than a predetermined threshold value. In this example, the non-link transition component does not transition to a node whose evaluation value is equal to or less than the threshold value.

  Further, according to the calculation rule of the time evolution calculation unit 18, it is assumed that the ratio of the link transition component in the evaluation value of the node increases monotonously as the evaluation value of the node increases. In this case, the ratio of the non-link transition component to the evaluation value of the node monotonously decreases as the evaluation value increases. By doing in this way, the evaluation value of a node with a low evaluation value tends to gather at a node with a high evaluation value instead of transitioning along a link. On the other hand, the evaluation value of a node having a high evaluation value tends to transition to another node along the link. Therefore, if there is a part where nodes with high evaluation values gather to some extent in the network and are connected to each other by links, even if the time evolution is repeated, the evaluation values circulate between the nodes of that part via the links. As a result, the evaluation value of the node in that part generally increases. This “part” forms one “community”. On the other hand, if a node with a high evaluation value is isolated at a position away from another node group with a high evaluation value in the network, the evaluation value flows out to the other node via the link, while the other node Since the evaluation value flowing into the node is small, the evaluation value of the node becomes lower as a result of time development. Therefore, even if there is a node having a high initial evaluation value obtained from the query, if the node is isolated from others, the node is difficult to be extracted as a member of the community.

  In one example, the link transition component is set to 0 for a node whose evaluation value is less than (or below) a certain threshold value θ. In this example, all evaluation values of nodes whose evaluation value is less than the threshold value θ are treated as non-link transition components, and as described above, priority is given to nodes with high evaluation values regardless of the link of the network. Transition to.

  The source of the evaluation value component that transitions is the initial evaluation value of each node obtained by the initial evaluation value calculation unit 16 based on the query, and the evaluation value of a node with a high initial evaluation value is mainly linked via a link. It flows out to the node. However, in a portion where a node group having a high initial evaluation value is closely connected by a link (that is, a community), an evaluation value component flows in through a link from another node having a high evaluation value. The evaluation value of the node is kept high. On the other hand, the evaluation value of the node goes down from that part and the non-link transition component becomes dominant, so the evaluation value returns to the node with a higher evaluation value than the transition to the link destination. There will be more parts to go. In particular, for a node whose evaluation value is less than the threshold θ, in the method in which the link transition component is set to 0, the evaluation value of the node whose threshold value is less than the threshold value does not flow out of the community via the link, and the total amount is the evaluation value. To higher nodes (mostly nodes in the community). Therefore, as time progresses, the evaluation value of the node group inside the community is kept high, while the evaluation value of all the nodes outside the community becomes zero. Therefore, if nodes whose final evaluation value after time evolution has converged are extracted, a set of these extracted nodes becomes a community corresponding to the initial query. Note that 0 used as the extraction threshold here is an ideal case. In practice, a node outside the community may have a small positive evaluation value even after convergence for some reason, so that the extraction threshold is set to a positive value so that such an external node is not extracted as a member of the community. A value (for example, the threshold value θ described above) is used.

As a specific example, the time evolution calculation unit 18 calculates the time evolution of the evaluation value p _n (t) of each node n using the following equations 1-3.

である。またＮはネットワークの全ノードの数である。すなわち、ノードの識別番号であるｎは１〜Ｎの値をとる。またθは評価値ｐ_n（ｔ）に対して設定された閾値である。 Here, p _n (t) is an evaluation value possessed by the node n at the time t, and is a value that has been normalized so that the sum of the evaluation values of the node n in the entire network is 1. That is,

It is. N is the number of all nodes in the network. That is, n, which is a node identification number, takes a value of 1 to N. Θ is a threshold set for the evaluation value p _n (t).

The function f _n (t) defined by Equation 2 is a link transition component in the evaluation value p _n (t) of the node n. In this example, the link transitions component f _n (t), the evaluation value is 0 if p _n (t) is less than the threshold value theta, if evaluation value p _n (t) is the threshold value theta higher (p _n ( t) −θ). An example of a graph showing the link transition component f _n (t) with respect to the evaluation value p _n (t) in this example is shown in FIG. In FIG. 2, the threshold θ is set to 0.1, but this is only an example.

The thick solid line graph shown in FIG. 2 indicates the link transition component f _n (t) according to Equation 2. For reference, the broken line graph shows the link transition when the evaluation value p _n (t) is entirely the link transition component f _n (t) (that is, f _n (t) = p _n (t)). The component f _n (t) is shown.

As shown in Equation 3, the function F (t) is the sum of the link transition components f _n (t) of all the nodes in the network.

The first term on the right side of Equation 1 represents an evaluation value component caused by the transition of the link transition component. This component is evaluated when the link transition component f _m (t−1) of each node m at time (t−1) transitions to the node n according to the transition matrix T _nm indicating the connection relationship between the nodes in the network. The sum of values.

  In addition, the second term on the right side of Equation 1 represents an evaluation value component caused by the transition of the non-link transition component. This component will be described in detail below.

The non-link transition component in the evaluation value of the node is expressed by the following equation.

When the non-link transition components are summed over all the nodes of the network, the following equation is obtained.

As can be seen, [1-F (t−1)] included in the second term on the right side of Equation 1 is the sum of the non-link transition components of all the nodes of the network at time (t−1). The second term on the right side is a value obtained by multiplying the sum by the ratio of the link transition component f _n (t−1) of the node n of interest to the sum. That is, the second term on the right side is obtained by distributing the sum of the non-link transition components of all nodes in proportion to the size of the link transition component of each node n. Therefore, in this example, the higher the evaluation value p _n (t) is, the more evaluation value distribution derived from the non-link transition component is received.

  In Formula 1, the length of a unit time, which is one cycle of time evolution, is set to 1 for the sake of brevity. Even when one cycle is set to other lengths, the calculation can be performed in the same way as in Equation 1.

In this example, the time evolution calculation unit 18 uses the set of initial evaluation values {p _n (0)} (time t = 0) of each node n calculated by the initial evaluation value calculation unit 16 as a starting point. The time evolution of is repeatedly calculated. A set of initial evaluation values {p _n (0)} as a starting point is called an initial pattern. The time evolution calculation unit 18 repeats the calculation of Expression 1 while advancing the time t by one unit time until the evaluation value p _n (t) converges. The determination of convergence may be performed by a conventional method. For example, for all nodes n, when the difference between the evaluation value p _n (t) at a certain time and the evaluation value p _n (t−1) at the previous time is equal to or less than a predetermined threshold value, the evaluation value p _n What is necessary is just to determine with (t) having converged.

The evaluation value p _n ^stead of the node n after convergence obtained in this way is the degree of conformity of the node n to the query from the user, taking into ^account the influence of the connection (link) relationship between the nodes in the network. Represents.

The community extraction unit 20 extracts a community (a set of nodes) that ^matches the query from the evaluation value p _n ^stead after convergence (that is, steady state) obtained by the time evolution calculation unit 18. For example, a node n whose steady state evaluation value p _n ^stead is larger than a predetermined threshold value (a value greater than or equal to 0) is extracted as a member of the community. Moreover, the community extraction part 20 may produce | generate the list of the members of the extracted community, and may display on a screen. This list may be obtained by sorting the members in the order of final evaluation values.

A set {p _n ^stead } of the evaluation values of the node n after convergence obtained as a result of the calculation of the time evolution calculation unit 18 is one of the attractors in a state that the set of evaluation values of the node n of the network can take. is there. Even if the initial pattern {p _n (0)} is slightly changed, the same attractor is reached by time evolution calculation. Therefore, even if the user does not specify the initial pattern strictly, the same attractor can be reached by the time evolution calculation if it can be specified with a certain degree of accuracy.

Even if the initial pattern {p _n (0)} includes a node that is located outside the desired community but has a high initial evaluation value, most of the nodes with a high initial evaluation value are included in the community. If it is included, it is possible to reach the same attractor in which the evaluation values of all the nodes outside the community are 0 to small values by the time evolution calculation. For example, it is assumed that the node 102 group inside the broken-line circle 110 in the network 100 shown in FIG. Here, it is assumed that the six dark-colored nodes in (a) are nodes having a high initial evaluation value obtained from the query input by the user. In this example, many of the nodes with high initial evaluation values are included in the community 110, but some are outside the community 110. These nodes with high evaluation values outside the community 110 are false members for the community 110. When the iterative calculation of the time evolution calculation unit 18 is applied to the initial pattern including such a false member, as shown in FIG. 3B, the node 102 group inside the community 110 has a high evaluation value (dark color). All of the nodes outside the community 110, including those that were false members, have an evaluation value of 0 to very small. That is, even if some of the false members are included in the initial pattern obtained from the query, the time evolution calculation unit 18 and the community extraction unit 20 extract a correct community that does not include the false members.

  As described above, if the query that the user first inputs can accurately specify the community to be obtained to some extent, the community to be obtained can be extracted. A node located outside the desired community but having a high initial evaluation value can be said to be a fake member from the viewpoint of the community, but in the method of this embodiment, even if a fake member is included in the initial pattern, You can extract the community you want.

  The embodiment described above is merely an example, and various modifications are possible within the scope of the present invention.

For example, the definition formula of the link transition component f _n (t) shown in the above formula 2 (see also FIG. 2) is merely an example. For example, as another example, if the evaluation value p _n (t) is less than the threshold value θ, f _n (t) = 0, but if it is greater than or _{equal to} θ, a definition formula that satisfies f _n (t) = p _n (t) may be used. Good. In this example, for nodes whose evaluation value p _n (t) is equal to or greater than the threshold θ, the entire amount of the evaluation value p _n (t) is a link transition component, and the non-link transition component is zero. The graph of the link transition component f _n (t) in this example is like the solid line graph shown in FIG.

In the example of FIG. 2 (Equation 2) and FIG. 4, the link transition component f _n (t) is 0 for the range where the evaluation value p _n (t) is less than the threshold θ, but this is also an example. Not too much. If it meets the condition that the proportion of the links in the evaluation value p _n (t) transition component f _n (t) is monotonically increases with an increase in the evaluation value p _n (t), the evaluation value p _n (t ) Between 0 and the threshold θ, the value of the link transition component f _n (t) is larger than 0 (for example, f _n (t) gradually increases from 0 as p _n (t) increases from 0, etc.) ). If this condition is satisfied, it will be difficult for a node in the community to flow out to a node outside the community, so that the difference in evaluation value between the community and the outside becomes clear due to time development. Incidentally, link transition component f _n (t), of the evaluation values p _n (t), but to indicate the amount of component a transition along the links emanating from the node n, evaluation values p _n (t) The value will never exceed.

  In addition, (a) a threshold value θ (see FIG. 2) indicating an upper limit evaluation value where the link transition component is 0, and (b) a threshold value which is a lower limit of the evaluation value that a node extracted as a community should have. And may not be the same value.

  The information processing apparatus exemplified above is realized by causing a computer to execute a program representing the function of each apparatus. Here, the computer includes, for example, a microprocessor such as a CPU, a memory (primary storage) such as a random access memory (RAM) and a read only memory (ROM), an HDD (hard disk drive), and an SSD (solid state drive) as hardware. ), Etc., a controller that controls a fixed storage device, various I / O (input / output) interfaces, a network interface that performs control for connection to a network such as a local area network, etc. It has a configuration. Also, portable non-volatile recording of various standards such as a disk drive and a flash memory for reading and / or writing to a portable disk recording medium such as a CD or a DVD via the I / O interface, for example. A memory reader / writer for reading from and / or writing to a medium may be connected. A program describing the processing contents of each functional module exemplified above is stored in a fixed storage device such as an HDD via a recording medium such as a CD or DVD, or via a communication means such as a network, and installed in a computer. Is done. The program stored in the fixed storage device is read into the RAM and executed by a microprocessor such as a CPU, thereby realizing the functional module group exemplified above.

  DESCRIPTION OF SYMBOLS 10 Network information storage part, 12 Network information acquisition part, 14 Query reception part, 16 Initial evaluation value calculation part, 18 Time development calculation part, 20 Community extraction part.

Claims (5)

Computer
An evaluation value calculating means for calculating an evaluation value indicating the degree of conformity to the query of each node in the network;
The first component of the evaluation value of each node transitions to any one of the linked nodes according to the link exiting from the node, and the second component of the evaluation value of each node is a link exiting from the node. Is a second component transition calculating means for calculating an evaluation value of each node after transition according to a calculation rule of transitioning to another node regardless of the second rule. Is more transitioned to nodes having an evaluation value higher than the first range than nodes having an evaluation value in the first range close to 0, and the first component occupying the evaluation value of the node The rate is monotonously increased as the evaluation value increases, transition calculation means,
A specifying means for specifying one or more nodes corresponding to the query based on the evaluation value after the transition of each node calculated by the transition calculating means;
Program to function as.   The specifying unit specifies, as a community corresponding to the query, a node group in which an evaluation value after transition of each node calculated by the transition calculating unit is larger than a predetermined threshold value. The program according to claim 1.   3. The list according to claim 1, wherein the specifying unit generates a list indicating the specified node group in descending order of evaluation values after transition of the nodes calculated by the transition calculating unit. program.   4. The calculation rule according to claim 1, wherein the ratio of the first component to the evaluation value of the node is 0 for a node whose evaluation value is equal to or less than a threshold value. 5. The program described in. An evaluation value calculating means for calculating an evaluation value indicating the degree of conformity to the query of each node in the network;
The first component of the evaluation value of each node transitions to any one of the linked nodes according to the link exiting from the node, and the second component of the evaluation value of each node is a link exiting from the node. Is a second component transition calculating means for calculating an evaluation value of each node after transition according to a calculation rule of transitioning to another node regardless of the second rule. Is more transitioned to nodes having an evaluation value higher than the first range than nodes having an evaluation value in the first range close to 0, and the first component occupying the evaluation value of the node The ratio monotonously increases as the evaluation value increases, and transition calculation means,
Identification means for identifying one or more nodes corresponding to the query based on the evaluation value after the transition of each node calculated by the transition calculation means;
An information processing apparatus.

Images